AI features#

Optional AI layer rename and visual grouping live in the plugin UI. They call any OpenAI-compatible chat API with vision support.

These features are independent of MCP. Bridge tools never need an LLM API key.

#Setup

1. Open the plugin → ⚙ → choose UI language (中文 / English) if needed
2. ⚙ → Model settings → set API base URL (default https://api.openai.com/v1)
3. Set Model (default gpt-4o; gpt-4o-mini also works)
4. Paste your API key
5. Click Test connection, then Save

Credentials and language preference are stored in Figma clientStorage on your machine only. UI strings live in packages/figma-agent-plugin/src/ui/locales.json.

#Prompt templates

⚙ → Prompt settings — edit system prompts for rename and group separately.

• Defaults: packages/figma-agent-plugin/src/prompts/*.prompt.txt
• Placeholders: {{candidates}}, {{suggestedClusters}} (group)
• Restore defaults reloads repo templates; Save writes overrides to clientStorage

#Custom AI providers

Figma plugins must declare allowed domains in manifest.json. This repo already allowlists common hosts, including:

For Azure OpenAI or other custom hosts: add the host to networkAccess.allowedDomains in manifest.json, rebuild, and re-import the plugin.

#AI rename

1. Select a single Frame or Group
2. Open the Rename tab → Start rename
3. The plugin clones the selection to the right, exports a 1× PNG, and sends layer candidates + image to your API
4. Returned names are applied on the clone (original is untouched)

Skipped: TEXT nodes and very small layers (< 2px).

#AI visual grouping

1. Select a single Frame, Group, or Section
2. Open the Group tab → Start group
3. Review/edit the JSON plan in the editor
4. Click Apply groups to create nested groups on the clone (depth-first)

The collector also suggests simple row-based proximity clusters as a hint to the model.

#Version updates

The plugin fetches:

https://raw.githubusercontent.com/ChinaCarlos/figma-agent-kit/main/releases/version.json

See Plugin release for how version.json is updated.